mpv-player
vo_gpu: revert 8a09299 and conditionally clear framebuffer by default
Just leaving my 2 cents. On x11 with amdgpu and opengl, I always see the garbled frames mentioned in that commit. Conditionally or always clearing the framebuffer seems to make no difference. With or without compositing on, switching fullscreen, resizing, etc. all leads to weird artifacts. This is with linux 5.9.10 and mesa 20.2.3 so who knows what changed since that commit was originally made.
Of course someone else may have a different findings but right now I don't see why we can't just revert it. There doesn't seem to be any benefit on my end from always clearing the framebuffer. @aufkrawall does the current mpv master behavior of always clearing the framebuffer still have improvements for you over conditionally clearing it?
Related issue: #7105
It hasn't changed since my last comment on it, 8a0929973de15d9574595c5a098bb3446757ef16 didn't help much. I still use radeonsi_zerovram in .drirc, which gets rid of it entirely (I think the env var for it has been removed recently). Imho a MR for Mesa, so that it does this by default, still seems like an attractive thing to do.
reopening this to start another discussion since the changes of the other accepted PR (#8372) were reverted (d9008d2aa84467064c3d76407f5f6a5b86e76520).
i didn't change anything yet to safe me from unnecessary work, eg i would like to first sort out some question. is this acceptable? if not, is anything acceptable and how should it be done? if the changes here are acceptable, what should be the default (since i guess the default should be always) and/or what should be changed?
what's your opinion on that @decent-name? i don't want to play revert ping pong with you.
okay if you are fine with the default like this (conditional) i am leaving it as is. also saw it doesn't work in dumb mode so i added it to the whitelist and rebased with current master.
if you are fine with the changes like this, it would be nice if you could approve the PR or merge it.